[FFmpeg-devel] N-59134-g1c67ad9 seems tobreakavformat_open_input, avformat_find_stream_info

[Don Moir]

----- Original Message ----- 
From: "Don Moir" <donmoir at comcast.net>
To: "FFmpeg development discussions and patches" <ffmpeg-devel at ffmpeg.org>
Sent: Monday, December 16, 2013 1:52 AM
Subject: Re: [FFmpeg-devel] N-59134-g1c67ad9 seems tobreakavformat_open_input, avformat_find_stream_info


> >On Mon, Dec 16, 2013 at 01:18:47AM -0500, Don Moir wrote:
>>> Just built  N-59134-g1c67ad9 from latest download and avformat_open_input and avformat_find_stream_info don't seem to work 
>>> anymore.
>>>
>>> I use windows.
>>>
>>> An mp3 passed avformat_open_input and failed in avformat_find_stream_info.
>>>
>>> A video file failed in avformat_open_input. These are files I play all the time and no file seems to work anymore.
>
>>fate is green, which includes windows and fate passes fine here
>>locally on mingw32 & 64
>>so i cannot reproduce the problem you describe
>>maybe you could provide more details and try a fresh checkout /
>>make distclean maybe
>>git bisect, error messages, which fate tests fail, ... may also be
>>usefull
>
> Just did a fresh download and build and same problem. Now N-59141-g48d39c8 (mingw32)
>
> Everything builds and links as expected.
>
> Some build flags I use:
>
> --disable-muxers
> --disable-avfilter
> --enable-shared
>
> The first av_log message I get for any file is:
>
> "Value 2147483648.000000 for parameter 'blocksize' out of range [1 - 2.14748e+009]"
>
> I don't get this message in previously working build and get something normal.
>
> I will see if I can come up with some more details on it.

In avio_read,  fill_buffer is called and then s->read_packet. In the failing build s->read_packet returns 0. In the working build, 
s->read_packet return 32768 which is the buffer size. So appears nothing ever gets read resulting in failure.

So need to go deeper than that, but posting so maybe someone has an answer.